05 / The Frontier
Biological Neural Computing
Pioneering research into synthetic biological intelligence. How living neurons are being integrated with digital systems to create a new paradigm of hybrid computing.
Cortical Labs & Synthetic Intelligence
Cortical Labs is a company at the forefront of biological computing, developing systems that interface lab-grown human and mouse neurons with digital computational environments.
This approach explores Synthetic Biological Intelligence (SBI). Unlike artificial intelligence (AI), which simulates neural networks using silicon chips, SBI utilizes actual biological neurons—harnessing their inherent computational efficiency, adaptability, and capacity for self-organization.
Silicon vs. Biology
- ArchitectureStatic (Silicon) vs. Plastic/Adaptive (Biology)
- Energy EfficiencyMegawatts (Supercomputers) vs. ~20 Watts (Brain)
- LearningBackpropagation vs. Active Inference
Why this is not just \"AI with cells\"
Conventional AI is built by designing mathematical architectures and then optimizing parameters with algorithms such as gradient descent on digital hardware. Biological neural computing starts from living tissue that already has intrinsic excitability, plasticity, and self-organizing dynamics.
That difference changes the engineering problem. Instead of simply scaling compute and data, researchers must maintain viable tissue, shape feedback, interpret emergent activity, and work with a substrate that is adaptive but also variable and fragile.
Platform thinking
Systems like DishBrain and CL1 are best understood as platforms for experimentation. They are valuable because they let researchers ask how living neural tissue behaves under controlled stimulation, adaptation, and task conditions, not because they are ready to replace general-purpose silicon computing.
The DishBrain System
A foundational experiment demonstrating that biological neural networks can exhibit sentient behavior in a simulated environment by learning to play the game Pong.
1. Culturing & Interfacing
Approximately 800,000 human or mouse neurons are cultured on top of high-density multi-electrode arrays (MEAs). These electrodes serve a dual purpose: they can both stimulate the neurons electrically (write) and record their spontaneous or evoked action potentials (read).
2. Closed-Loop Feedback
The MEA acts as a bidirectional interface. In the Pong experiment, sensory information (the location of the ball) is encoded into electrical stimuli delivered to specific sensory regions of the culture. The culture's spontaneous electrical output is decoded as motor commands (moving the paddle).
3. Learning via Active Inference
Learning is induced not by traditional reward mechanisms, but by the Free Energy Principle. The system receives structured, predictable stimulus when it hits the ball, and chaotic, unpredictable "white noise" stimulus when it misses. Seeking predictability, the neural network self-organizes its activity to keep the ball in play, effectively learning the task within minutes.
System Flow Diagram
Critical Analysis & Limitations
Experimental vs. Practical Deployment
It is vital to distinguish between proof-of-concept experimental research and commercially viable technology. DishBrain demonstrates rudimentary learning in a highly controlled, simplified environment. We are decades away from biological chips replacing silicon in laptops or data centers.
Scalability & Stability
Biological systems are inherently fragile. Culturing neurons requires precise environmental controls (temperature, nutrients, CO2). Scaling up from 800,000 neurons to the millions or billions required for complex computation poses massive biochemical engineering challenges.
Reproducibility Challenges
Unlike identical silicon chips, every neural culture is unique. The stochastic nature of biology means two cultures may form different synaptic connections and solve identical tasks using entirely different network topologies, complicating standardized computation.
Ethical Considerations
The use of living human neurons for computation raises unprecedented bioethical questions. As these systems scale, determining the moral status of complex in-vitro networks, addressing the definitions of sentience vs. consciousness, and regulating synthetic biological intelligence will require robust ethical frameworks before widespread deployment.
How to read claims in this field
Biological computing headlines can easily become exaggerated. A careful reader should ask what the system actually did, how the task was defined, what kind of feedback it received, how learning was measured, and whether the authors are describing a proof of concept, a platform, or a deployable product.
Those questions help separate real advances from speculative narratives and keep attention focused on what the experiments genuinely demonstrate.
Potential value of the field
Even if biological computers never become mainstream replacements for digital devices, the field could still be highly valuable. These platforms may improve disease modeling, drug testing, and our understanding of how neural systems learn under embodied feedback.
In that sense, neural computing research matters not only for hardware innovation but also for basic neuroscience and translational biomedicine.
Resources Used
These references support the educational summaries on this page and are included so readers can continue into primary or institutionally reviewed material.
In vitro neurons learn and exhibit sentience when embodied in a simulated game-world
Primary source for DishBrain, closed-loop feedback, and learning in cultured neurons.
The technology, opportunities, and challenges of Synthetic Biological Intelligence
Used for the field-level overview of SBI opportunities, limits, and ethical challenges.
CL1
Provides the primary product and platform description for current commercial biological computing efforts.
The free-energy principle: a unified brain theory?
Supports the explanation of active inference as a conceptual framework used in DishBrain discussions.